Knowledge is Believing Something Because It’s True
Penultimate draft. Please cite final draft, forthcoming in Episteme.
Tomas Bogardus
Will Perrin
Pepperdine University
Northern Illinois University
tomas.bogardus@pepperdine.edu
wperrin@niu.edu
Abstract
Modalists think that knowledge requires forming your belief in a “modally stable” way: using a
method that wouldn’t easily go wrong (i.e. safety), or using a method that wouldn’t have given
you this belief had it been false (i.e. sensitivity). Recent Modalist projects from Justin ClarkeDoane and Dan Baras defend a principle they call “Modal Security,” roughly: if evidence
undermines your belief, then it must give you a reason to doubt the safety or sensitivity of your
belief. Another recent Modalist project from Carlotta Pavese and Bob Beddor defends “Modal
Virtue Epistemology”: knowledge is a belief that is maximally modally robust across “normal”
worlds. We’ll offer new objections to these recent Modalist projects. We will then argue for a
rival view, Explanationism: knowing something is believing it because it’s true. We will show how
Explanationism offers a better account of undermining defeaters than Modalism, and a better
account of knowledge.
Introduction
The analysis of knowledge: what all the ages have longed for.1 If you’re an epistemologist, at least.
Though the long and fruitless search has jaded many of us, we present here what we believe is the
genuine article. To convince you, we’d like to discuss an emerging rivalry between two opposing views
of knowledge, which we’ll call “Modalism” and “Explanationism.” We’ll raise objections to Modalism,
and then argue for Explanationism. According to Modalism, knowledge requires that our beliefs track
the truth across some appropriate set of possible worlds.2 Modalists have focused on two modal
1
Our topic is propositional knowledge, i.e. knowledge-that. We’ll leave this qualification implicit in what follows.
For examples of Modalist projects, see Tim Black (2008), Tim Black and Peter Murphy (2007), Justin Clarke-Doane
(2012, 2014, 2015, 2016), Keith DeRose (1995), Fred Dretske (1971), Jonathan Jenkins Ichikawa (2011), Steven
Luper-Foy (1984), Robert Nozick (1981), Duncan Pritchard (2007, 2009), Sherrilyn Roush (2005), Mark Sainsbury
(1997), Ernest Sosa (1999a, 1999b, 2007, 2009), and Timothy Williamson (2002).
2
1
conditions. Sensitivity concerns whether your beliefs would have tracked the truth had it been
different.3 A rough test for sensitivity asks, “If your belief had been false, would you still have held it?”
Safety, on the other hand, concerns whether our beliefs track the truth at ‘nearby’ (i.e. ‘similar’) worlds.4
A rough test for safety asks, “If you were to form your belief in this way, would you believe truly?”
Modalism was a natural outgrowth of the revolutionary work done in modal logic during the last
century. When Kripke, Lewis, and the gang give you a shiny new hammer, everything starts looking like a
nail. Even knowledge.
According to Explanationism, by contrast, knowledge requires only that beliefs bear the right
sort of explanatory relation to the truth. In slogan form: knowledge is believing something because it’s
true. 5 Less roughly, Explanationism says knowledge requires only that truth play a crucial role in the
explanation of your belief.6 What’s a crucial role? We propose adapting Michael Strevens’ (2011)
“kairetic test” for difference-making in scientific explanation, to the more general purposes of
Explanationism. With regard to scientific explanation, Strevens’ proposal is this: Start with a deductive
argument, with premises correctly representing some set of influences (potential explanans), and the
conclusion correctly representing the explanandum. Make this argument as abstract as possible, while
preserving the validity of the inference from premises to conclusion; strip away, as it were, unnecessary
information in the premises. When further abstraction would compromise the validity of the inference,
stop the process. What’s left in the premises are difference-makers, factors that play a “crucial role” in
the scientific explanation. Now, explaining why a belief is held is importantly different from paradigm
cases of scientific explanation. In order to adapt the kairetic test to Explanationism, we propose
beginning with a set of potential explanans that explain the relevant belief’s being held, and then
proceeding with the abstraction process until further abstraction would make the explanation fail. The
remaining explanans are the difference-makers. On Explanationism, for the belief to count as
3
A more precise statement of the sensitivity condition would specify the method being used. For examples of
sensitivity theorists, see Fred Dretske (1970), Robert Nozick (1981), and Kelly Becker and Tim Black (2012).
4
For examples of safety theorists, see Duncan Pritchard (2005), Ernest Sosa (1999), John Hawthorne (2004), and
Mark Sainsbury (1997).
5
Compared with many other analyses of knowledge on offer, this one is suspiciously simple. But we think Aristotle
was right that we must not expect more precision than the subject matter admits. Also, we find it charming that
Explanationism proposes to analyze as ancient and august a concept as knowledge in terms no less venerable:
explanation, belief, and truth. On top of its philosophical merits, this is historically and psychologically plausible.
For examples of broadly Explanationist projects, see Alan Goldman (1984), Steven Rieber (1998), Ram Neta (2002),
Marc Alspector-Kelly (2006), Carrie Jenkins (2006), Dan Korman and Dustin Locke (2018), and David Faraci (2019).
6
As Alan Goldman put it, the truth of your belief “enters prominently” into the best explanation for its being held.
2
knowledge, the truth of the relevant belief must be among these difference-makers in order a belief to
count as knowledge.
We believe that Explanationism’s time has come. The iron is hot; the field is white for the sickle.
Below, we will raise objections to two recent Modalist projects. The first project comes from Justin
Clarke-Doane and Dan Baras (forthcoming), and centers on a principle they call “Modal Security.” The
second project is a defense of “Modal Virtue Epistemology,” by Bob Beddor and Carlotta Pavese
(forthcoming). Our criticisms of these two projects point toward the shortcomings of Modalism
generally. We’ll then set Modalism aside, and move to a motivation and defense of Explanationism.
We’ll let the view stretch its legs in straightforward cases of perception, memory, introspection, and
intuition. Then, we’ll put Explanationism through its paces, with harder cases of abduction, deduction,
and induction, as well as a couple types of apparent counterexamples that might occur to the reader.
We’ll close by showing how Explanationism defuses Linda Zagzebski’s Paradox of the Analysis of
Knowledge.
Modal Security
Justin Clarke-Doane, in his earlier work (2015 and forthcoming, §4.6), as well as in more recent work
with Dan Baras (forthcoming, 1), defends the following principle:
Modal Security: If evidence, E, undermines (rather than rebuts) my belief that P, then E gives
me reason to doubt that my belief is sensitive or safe (using the method that I actually used to
determine whether P).
Though this principle does not mention knowledge, we count it as part of a Modalist project because of
the connection between undermining defeaters and knowledge. Baras and Clarke-Doane say that, in
contrast to a rebutting defeater, an undermining defeater does not target the truth of a belief, and so “it
must show that there is some important epistemic feature that our belief that P lacks. Safety and
sensitivity, along with truth, have emerged from recent epistemology literature as among the most
important epistemological features of beliefs.” So, if neither Safety nor Sensitivity is required for
knowledge, presumably they’re not “important epistemic features,” the lack of which would allow for an
undermining defeater, and doubts about which must be produced by any undermining defeater. And
that would be trouble Modalism generally, and for Modal Security in particular.
3
Baras and Clarke-Doane respond to several proposed counterexamples in the literature. For
example, David Faraci (2019) proposes Eula’s Random Number Generator. Eula has good (but
misleading) reason to believe her computer is running a program designed to produce prime numbers.
Actually, it’s merely a random number generator. But, by chance it happens to produce only primes.
And, as Baras and Clarke-Doane say, “these events are maximally deterministic. At every
metaphysically—not just physically—possible world,” Eula gets the same results.7 So, for any number n
that the computer spits out, if Eula believes that n is prime, her belief is formed safely. (And
sensitively).8 Despite this safety (and sensitivity), the objection concludes, Eula get a defeater for her
belief when learns she’s dealing with a random number generator, one that doesn’t produce these
outputs because they’re prime. This is trouble for Modal Security, which says that defeaters must come
by way of safety or sensitivity violations.
Baras and Clarke-Doane notice that the problem with this case is in the idea of a number
generator that is both random, but which must—or even merely which would—deliver only primes. We
put the problem this way: If the number generator is truly random, it can’t be the case that it must (or
would) produce only primes. This is a problem, since we judge that Eula gets a defeater in this case only
insofar as the number generator is genuinely random. But we judge that the number generator is a safe
(and sensitive) method of forming beliefs about prime numbers only insofar as it’s not random, but must
(or would) produce only primes. In other words, Faraci needs a case in which a subject gets a defeater
for a safely- and sensitively-formed belief. But if Eula’s number generator is truly random, then it’s
neither a safe nor a sensitive method. Yet if it’s not truly random, and for some reason reliably produces
primes, then Eula gets no defeater when she learns this.
We would like to present a new problem for Modal Security. Recall that Modal Security says that
if evidence undermines your belief, it must give you reason to doubt that your belief is sensitive or safe.
And this is because either sensitivity or safety is an “important epistemic feature,” which we take to
7
In his own words, Faraci (2019, 22) says of this case, “At every possible world, Eula’s counterpart forms beliefs
about which numbers are prime by consulting the Source’s counterpart, and at every possible world, the Source’s
counterpart delivers the same answers as at the actual world. [There is no explanatory connection between this
modal stability and the facts about which numbers are prime.]” But he adds, in a footnote, that all the argument
really needs is that the number generator spit out only prime numbers in all the “nearby” worlds, so that the
safety counterfactual comes out true: if Eula were to from a belief about primes using this method, she would
believe truly. As long as the number generator doesn’t deliver these numbers because they’re prime, Faraci thinks,
learning that it’s a random number generator should give her a defeater, even without targeting safety.
8
If her belief were false—that is, if per impossibile this number weren’t prime—then the computer would not have
produced it as an output, and so she would not have believed that it was prime.
4
mean a necessary condition on knowledge. For if neither safety nor sensitivity is required for knowledge,
it’s hard to see how an undermining defeater could possibly succeed by targeting either of these
conditions, and harder still to see why an undermining defeater must provide reason to doubt that these
conditions obtain. If some condition is not necessary for knowledge—if, that is, if one can transgress this
condition and still know full well that p is true—then why think evidence that this condition isn’t met
should defeat one’s belief that p? And if neither sensitivity nor safety is required for knowledge, then
that means some other condition is, in addition to truth and belief, since truth and belief are themselves
insufficient for knowledge. And, presumably, an undermining defeater could succeed by targeting this
other condition, without targeting sensitivity or safety. If these additional assumptions about the
relationship between undermining defeat and knowledge are right, then a case of belief that was not
formed sensitively or safety and yet is knowledge all the same would refute Modal Security.
We believe there are such cases, and here’s a general recipe to whip one up: first, pick the most
virtuous belief-forming method you can imagine, and have a subject form a belief via that method.
Second, add a twist of fate: put the method in danger of malfunctioning, but let the danger remain
purely counterfactual. Now, since things could have gone less well epistemically but didn’t, it is quite
tempting to admit that the virtuous method produced knowledge. And yet the belief was not formed
safely, for there are many nearby possible worlds in which the method goes awry. Such a case nicely
pries apart our concept of knowledge from our concept of safety, by polluting very many of the
“nearby” worlds with false beliefs, while maintaining a tight enough connection between belief and its
truth to allow for knowledge.9 What we see in such cases is that, sometimes, our actual performance is
enough for knowledge, even under threat of externalities.
For example, suppose that the world’s most accurate clock hangs in Smith’s office, and Smith
knows this. This is an Atomic Clock: its accuracy is due to a clever radiation sensor, which keeps time by
detecting the transition between two energy levels in cesium-133 atoms. This radiation sensor is very
sensitive, however, and could easily malfunction if a radioactive isotope were to decay in the vicinity.
This morning, against the odds, someone did in fact leave a small amount of a radioactive isotope near
the world’s most accurate clock in Smith’s office. This alien isotope has a relatively short half-life, but–
quite improbably–it has not yet decayed at all. It is 8:20 am. The alien isotope will decay at any moment,
but it is indeterminate when exactly it will decay. Whenever it does, it will disrupt the clock’s sensor,
9
According to Explanationism, this “tight enough” connection is the explanation relation. In Atomic Clock, it
remains the case that Smith believes it’s 8:22 because it is 8:22, even if, due to the threat of this isotope, the actual
world is surrounded by an ocean of error.
5
and—for fascinating science-y reasons that needn’t delay us here—it will freeze the clock on the reading
“8:22.”
The clock is running normally at 8:22 am when Smith enters her office. Smith takes a good hard
look at the world’s most accurate clock—what she knows is an extremely well-designed clock that has
never been tampered with—and forms the true belief that it is 8:22 am. Now, does Smith know that it’s
8:22 am? It strikes us as obvious that she does, and perhaps the same goes for you. Like Harry
Frankfurt’s (1969, 835-6) counterfactual intervener, the damage of this radioactive isotope remains
purely counterfactual, and everything is functioning properly when she forms her belief. Yet, since the
isotope could easily have decayed and frozen the clock at “8:22”, and since Smith may easily have
checked the clock a moment earlier or later, Smith might easily have believed it is 8:22 am without it is
being 8:22 am. Smith formed her belief in a way that could easily have delivered error. So, while the
actual world remains untouched, the threat this isotope poses to the clock pollutes “nearby” modal
space with error. This allows us to conceptually separate knowledge from safety.
What’s more, Smith’s belief is not formed sensitively. Recall the test for sensitivity: If it weren’t
8:22, Smith wouldn’t believe that it is (via this method). Let’s run the test. How would things be for Smith
and her clock if it weren’t 8:22? Well, if it weren’t 8:22 when Smith checked her clock, it would be
slightly earlier or later. And the isotope has been overdue to decay for some time now. So, if Smith were
to check the clock slightly earlier or later, she may well check a broken clock: the isotope may well have
decayed, freezing the clock on “8:22” even when it’s not 8:22. In that case, Smith would still believe it’s
8:22 by checking her clock. So, if it weren’t 8:22, Smith may well still believe that it is, using the same
method she actually used. So, it’s not the case that Smith would not believe it’s 8:22 by checking this
clock, if it weren’t 8:22. So, Smith fails the sensitivity condition.
Therefore, neither safety nor sensitivity is a necessary condition on knowledge. So, if
undermining defeaters must operate by targeting a necessary condition on knowledge, then
undermining defeaters can’t operate via safety or sensitivity, and a fortiori they need not so operate.10
10
Atomic Clock targets the alleged safety and sensitivity conditions on knowledge. Yet, as we’ve noted, Modal
Security is a principle that concerns only undermining defeat. If the additional assumptions connecting
undermining defeat to knowledge discussed above give you pause for thought, you may be pleased to learn that
Atomic Clock can be modified to target Modal Security directly. In Atomic Clock, there is no safety or sensitivity,
but there is knowledge. So, if Smith learns all this, she doesn’t get a defeater. But now suppose that the isotope
does decay and the clock goes out. It freezes. Learning that does give Smith an undermining defeater for her belief.
But, since the clock remains an unsafe (and insensitive) way to form a belief, the defeater doesn’t give Smith a
reason to think the clock is unsafe or insensitive. (She was already sure of that.) Or, if you think this still does give
her reason to doubt safety or sensitivity, even more, we could add another decaying isotope that would flip the
clock on, but this decay is really unlikely. We can adjust the probabilities so that the clock is exactly as unsafe and
6
Therefore, Modal Security is false.11 This ends our discussion of the first Modalist project we mean to
address. We turn now to the second Modalist project: Modal Virtue Epistemology.
Modal Virtue Epistemology
In a recent paper, Bob Beddor and Carlotta Pavese (forthcoming) endorse a version of the safety
condition on knowledge, and they offer this analysis: “a belief amounts to knowledge if and only if it is
maximally broadly skillful — i.e., it is maximally modally robust across worlds where conditions are at
least as normal for believing in that way.” For a performance to be maximally modally robust is for it to
succeed in all of the relevantly close worlds (ibid., 10). This is why we don’t need a truth condition in
their analysis: maximal modal robustness entails truth. As for “normal,” they provide this guide for
insensitive as before. There are exactly the same proportions of “nearby” worlds with a working clock and worlds a
broken clock. Since learning that the clock has frozen in this context would undermine Smith’s belief without
requiring any change in her confidence regarding safety or sensitivity, this is a problem for Modal Security. Notice
also that this case makes trouble even for “Modal Security3,” from Clarke-Doane and Baras (forthcoming, §8),
which says: “If evidence, E, undermines our belief that P, then E gives us direct reason to doubt that our belief is
sensitive or safe or E undermines the belief that <the belief that P is safe and sensitive>.” For reasons already
given, the first disjunct of the consequent of Modal Security 3 is not met. And, plausibly, neither is the second
disjunct, for two reasons: First, in the case now under consideration, Smith doesn’t even believe that <the belief
that it’s 8:22 is safe and sensitive>. In fact, since Smith had already been informed of the nearby isotope, Smith
already knew that the belief that it’s 8:22 is neither sensitive nor safe. Second, even if we read the second disjunct
of the consequent of Modal Security3 in such a way that Smith would need only receive reason to doubt (i.e. lower
her credence) in the proposition that <the belief that it’s 8:22 is safe or sensitive>, given the facts about the
“offsetting” isotope we describe above, plausibly Smith doesn’t receive any such reason.
11
Going by its letter and not its spirit, Modal Security entails merely this: either safety or sensitivity is itself
required for knowledge, or either safety or sensitivity entails some other feature F that is required for knowledge
(so that, if an undermining defeater operates by showing that your belief lacks F, it thereby also gives you a reason
to think your belief lacks sensitivity or safety, en passant). Atomic Clock deals with the first possibility, showing
that neither safety nor sensitivity is required for knowledge. As for the second possibility, neither believing
sensitively nor believing safely by itself entails knowledge, because neither one entails truth. Nor do they entail
truth even conjoined. And, as Linda Zagzebski (1994) taught us, if the non-true-belief part of a proposed analysis of
knowledge doesn’t entail truth, then it should be possible to construct Gettier cases for the analysis. And one can
do that for safety and sensitivity like so: pick the most virtuous, safest, most sensitive method you can think of.
Like a clock, or a calculator. It’s a safe method, so there are no ‘nearby’ worlds in which you use it and it delivers
error—but there are ‘distant’ worlds where this happens! It’s a sensitive method, so, there are no ‘nearby’ worlds
in which a proposition is false and yet the method tells you that proposition is true—but there are ‘distant’ worlds
where this happens! Now let a person employ this method to form a belief. Suppose this was one of the rare,
‘distant’ occasions when the method got it wrong. But, by sheer coincidence, events transpire that cause the belief
to turn out true. A Gettier case. So, the belief will not be knowledge, even though it’s a true belief formed via a
method that was both sensitive and safe. So, safety and sensitivity are not enough to entail whatever’s missing
from the analysis of knowledge in order to avoid Gettier cases. This shows that there is some anti-Gettier condition
that’s required for knowledge, yet not entailed by sensitivity or safety (or both). But then Baras and Clarke-Doane
should agree that an undermining defeater might succeed without giving us reason to doubt safety or sensitivity,
by instead giving us reason to doubt that this missing, anti-Gettier condition has been satisfied, where this
condition is not entailed by sensitivity or safety (or both). And this means Modal Security is false.
7
understanding the notion: “identify the normal conditions for a task with those which we would
consider to be fair for performing and assessing the task… [O]ur intuitions about cases reveal a tacit
grasp of these conditions.”
Beddor and Pavese offer a general strategy for responding to challenges to the safety condition,
and here’s how it would go with respect to Atomic Clock. The task at hand is: forming beliefs about the
time based on checking a clock. If the isotope had interfered with the clock, this interference would
have rendered conditions significantly more abnormal for the task at hand than they are in the actual
world, where the isotope is present but does not intervene. The decayed isotope puts poor Smith at an
unfair disadvantage, as it were. According to Modal Virtue Epistemology, we should ignore situations
like that, where the subject is disadvantaged by abnormal, unfair conditions. So, it turns out that Smith’s
belief is safe (i.e. maximally broadly skillful), since it is true in all worlds where conditions are at least as
normal as they are in the actual world. This how Beddor and Pavese mean to respond to proposed
counterexamples to Modal Virtue Epistemology.
However, there is a serious problem for this Modalist strategy from Beddor and Pavese. On their
view, knowledge requires safety, and their understanding of safety requires maximal modal robustness.
It’s a high-octane brand of safety that they’re using in this machine, requiring success in all relevantly
close worlds, i.e. worlds where conditions are at least as normal for the task. But what if we’re excellent
yet not infallible at some task? Remembering what we ate for breakfast, for example. We’re really good
at this. But sometimes we get it wrong, even in normal, fair conditions. Modal Virtue Epistemology says
that our belief was not maximally modally robust, and so not safe, and so not knowledge. On their view,
knowledge requires infallibility in all scenarios at least as normal or fair as the actual scenario. So
ordinary, fallible knowledge is a counterexample to Modal Virtue Epistemology. This is a problem.
But let’s be charitable, even if we fail in all other virtues. Suppose Pavese and Beddor retreat to
this—safety requires just substantial modal robustness, truth in many relevant nearby worlds. In that
case, we get the second horn of the dilemma. Now the view is: a belief amounts to knowledge if and
only if it is broadly skillful to a high degree. But, even still, there are counterexamples. Suppose that,
very improbably and acting totally out of character, we give you some eye drops that makes your vision
blurry, and then pressure you to read an eye chart. Your vision is so blurry that either you get the wrong
answer, or you get the right answer by chance. Neither one is knowledge. But what does this revised
view say? Well, we’ll have to check all the ‘nearby’ worlds at least as normal as the actual world, which
will include all the many worlds where, acting in character, we don’t administer the eye drops, and you
have clear eyes. In many of these worlds, you get the right answer. So, the belief is safe, on this revised
8
view. So, you have substantial modal robustness; the belief is substantially broadly skillful. So, it’s
knowledge, on this view. But this is certainly the wrong result.
We have good reason, then, to abandon Modal Virtue Epistemology, just as we did with Modal
Security. The failures of these two prominent Modalist projects are suggestive of deep flaws within
Modalism itself. We believe Modalism was an understandable application of exciting twentieth-century
advances in modal logic to longstanding problems in epistemology. And it comes close to the truth. But
it’s possible, albeit difficult, to design counterexamples that tease apart knowledge from sensitivity and
safety. So, we believe Modalism misses something crucial about the nature of knowledge: the
connection between a believer and the truth can’t be fully captured in modal terms, because it’s an
explanatory connection. Modalism, then, can have no long-term success as a research project. It’s
standing in its own grave. So, it’s time to look elsewhere for the analysis of knowledge.
Explanationism
What is Explanationism? Alan Goldman (1984, 101) puts it this way: a belief counts as knowledge when
appeal to the truth of the belief enters prominently into the best explanation for its being held. And
Carrie Jenkins (2006, 139) offers this definition: S knows that p if and only if p is a good explanation of
why S believes that p, for someone not acquainted with the particular details of S’s situation (an
‘outsider’).12 As for what makes something a good explanation, we recommend that Explanationism be
open-minded, and commit itself only to this minimal understanding of explanation from Goldman (ibid.):
“Our intuitive notion of explaining in this context is that of rendering a fact or event intelligible by
showing why it should have been expected to obtain or occur.”13 As for what it is for truth to “enter
12
Jenkins (2006, 138) says, “I present this condition, however, not as an analysis of knowledge, but rather as a way
of getting a handle on the concept and further the effort to understand what its role in our lives might be.” We,
however, advocate Explanation as a good old-fashioned analysis of knowledge. If you have Williamsonian scruples
against the project of analyzing knowledge, please see Faraci (2019, 22-3), who helps dispel those worries.
13
Goldman’s talk of “rendering a fact or event intelligible” may occasion the worry that, since what counts as
demystifying varies from subject to subject, on Explanationism, whether someone’s belief that p counts as
knowledge will vary from subject to subject. Some accounts of explanation—known as “pragmatic” or “contextual”
accounts—may well have this implication. See, for example, Bas van Fraassen (1980) and Peter Achinstein (1983).
But traditional, non-contextual accounts of explanation can give a different, objective reading of “rendering facts
intelligible” or “demystifying phenomena.” An explanation may succeed, render intelligible a phenomenon, and
demystify it, even if someone disagrees or fails to appreciate this. Such an account may also assuage the worry
that explanation itself is ultimately analyzed in terms of knowledge, perhaps by way of understanding. In which
case, by analyzing knowledge in terms of explanation, Explanationism will go in a circle. But if explanation is a
mind-independent relation, not to be analyzed in terms of mental states like understanding or knowledge, then
there is hope for Explanationism to avoid this circle. Though we ourselves favor traditional, non-contextual
accounts of explanation, we intend Explanationism to be consistent with each.
9
prominently” or, as we’d say, to figure crucially into an explanation, we remind our reader of what we
said above regarding Strevens’ kairetic test for difference makers: the most abstract version of the
explanation will feature all and only the difference-makers. If the truth of the relevant belief is among
them, that truth figures crucially into the explanation, and the relevant belief counts as knowledge.
One motivation for Explanationism is the observation that debunking arguments across
philosophy—against moral realism, religious belief, color realism, mathematical Platonism, dualist
intuitions in the philosophy of mind, and so on—have a certain commonality.14 To undermine some
belief, philosophers often begin something like this: “You just believe that because…,” and then they
continue by citing an explanation that does not feature the truth of this belief. Our evident faith in the
power of such considerations to undermine a belief suggests that we take knowledge to require that a
belief be held because it’s true, and not for some other reason independent of truth. Explanationism
agrees.
We’ll now further motivate Explanationism by taking the reader on a test drive through some
common avenues of gaining knowledge, to see how Explanationism handles. Let’s start with perception,
using vision as a paradigm example. Consider an ordinary case, for example the belief that there’s a
computer in front of you, formed on the basis of visual perception under normal conditions. Your
believing this is straightforwardly explained by the fact that there is a computer in front of you. It’s true
that there’s more we could (and perhaps even should) add to the explanation—facts about the lighting,
distance, functioning of your visual system, your visual experience, etc.—but the truth of your belief
would remain a crucial part of this explanation. If we tried to end the explanation like so, “You believe
there’s a computer before you because it looks that way,” we’d be ending the explanation on a
cliffhanger, as it were. We should wonder whether things look that way because they are that way, or
One may also worry that, if explanation itself is eventually cashed out in modal terms, then
Explanationism is not a viable alternative to Modalism at all, and in fact reduces to a species of Modalism. One
might think, for example, that if A explains B, then if A hadn’t obtained, B would not have obtained either, and if A
explains B, then if B were to obtain, then A would obtain as well. And the superficial similarity of such
implications—with respect to Explanationism’s claim that truth (A) must explain belief (B)—to the Sensitivity and
Safety conditions might make one uncomfortable. However, we think both of those inferences above are invalid,
subject to refutation by counterexample (e.g. cases familiar from the causation literature, featuring preemption,
trumping, etc.). Also, if we’re right that neither Sensitivity nor Safety are required for knowledge, and yet the
condition laid down by Explanationism is required for knowledge, then this shows that the condition laid down by
Explanationism does not entail either Sensitivity or Safety. While we’re optimistic that explanation need not be
analyzed in modal terms at all, even if that turned out to be the case, the fact that the condition laid down by
Explanationism is distinct from both Safety and Sensitivity would still count as an important advance. (We thank
our audience at the Institute of Philosophy in London for helpful discussion of this worry.)
14
See Dan Korman (2019) for a nice overview of this style of argument across the entire landscape of philosophy.
10
for some other reason.15 A satisfying, complete explanation in this case, then, will include the fact that
there is a computer before you, which is the truth of the relevant belief. So, Explanationism tells us this
is a case of knowledge, as it should.
Knowledge by way of memory can receive a similar treatment. Memory is, after all, something
like delayed or displaced perception. When you remember that you had granola for breakfast, you hold
that belief on the basis of a familiar kind of memorial experience. And, if this really is a case of
knowledge, you’re having this memory because you actually had the experience of eating granola for
breakfast, and you had the experience of eating granola because you really did eat granola. This last link
is required for a complete explanation. So, again, though there is much we might add to the explanation
of your belief in this case to make it more informative, the truth of your belief will play a crucial role.
And so Explanationism again gives the right result.16
We believe that Explanationism also handles introspection rather easily. Even if you don’t think
that introspection operates via representations, as perception does, probably we can agree that some
sort of spectating occurs during introspection. And so a story can be told much like the story above
about ordinary visual perception. When you believe, for example, that you’re desiring coffee, or that
you’re having an experience as of your hands sweating from caffeine withdrawals, you believe this
because of what you “see” when you attend to the contents of your own mind (or what you “just see,” if
you insist, rightly, that introspection is direct and non-representational). And you (just) see what you do
because it’s there. You really do have the desire, and you really do have the experience as of your hands
sweating. So, here too we find that truth plays a crucial role in the explanation of your belief. And
Explanationism gets the right result again.
Now let’s consider rational intuition. Here, things might look more challenging for
Explanationism. Roger White (2010: 582-83) puts the concern like so: “It is hard to see how necessary
15
It would be like this: A man of modest means presents his wife with a letter he’s just opened containing a
$50,000 check. “Where did this come from?” she excitedly and reasonably asks. “The mailbox,” he truthfully
replies. This is not a complete explanation. So too if we offer how things look to explain how we believe things are.
16
As for knowledge of events that happened in the past but which you don’t remember, Alan Goldman (1984, 103)
says something we consider very sensible: “On one variation a past fact explains subsequent perceptions and
beliefs that, together with communication, explain a present belief about that fact. On a second variation a past
fact leaves material traces (that it explains), which constitute present evidence that, together with beliefs about
the explanatory relations themselves, explain present belief about the past fact. Knowledge about the future
derives from similar chains as well. Presently observed facts can help to explain some belief about the future, and
these facts can also explain via causal chains the coming to pass of the future events now believed forthcoming on
the basis of this evidence.”
11
truths, especially of a philosophical sort can explain contingent facts like why I believe as I do.” But we
believe Explanationism has an answer. In the ordinary cases of knowledge of necessary truths, rational
intuition puts us in a position to “just see” the truth of the proposition. Emphasis on “just,” since, in this
case, no premises or inferences are needed. It’s akin to the case in which we “just see” that there’s a
computer in front of us, non-metaphorically, using our eyes. (Though, in the case of perception, there is
an appearance/reality distinction.) To the degree that it’s plausible that, in the ordinary case, you
believe there’s a computer before you because there is a computer before you, it’s also plausible that,
for example, you believe that 2+2=4 because, lo, this fact is there, before your mind’s eye, as it were.
Rational intuition has positioned you to appreciate this truth. The fact that 2+2=4 is necessarily true is
no barrier to its contingent presence before your mind playing a crucial role in the explanation of your
belief.17
17
An anonymous referee helpfully expressed this worry: the fact that 2+2=4 is causally inert, and so the fact is
literally not doing any work here. If there were no numbers, or if the number 2 bore the plus relation to itself and
to 7, we still would have believed as we do. In response, we deny that if there were no numbers, or if 2+2=7, we
still would have believed that 2+2=4. We believe that rational intuition puts us in direct contact with the fact that
2+2=4. Rational intuition, on our view, does not operate via any sort of report, experience, representation, or
appearance (e.g. what some call an “intellectual seeming,” cf. Bealer 1998), and so does not require that 2+2=4
have any causal power. On our view, there is no appearance/reality distinction when it comes to rational intuition:
our grasp or apprehension of the relevant fact is unmediated, a direct acquaintance. So, there is no opportunity for
hallucination or illusion when it comes to rational intuition. And, therefore, if it were false that 2+2=4, we would
have noticed that. Followers of physicalism may find it difficult to explain the direct apprehension of truth afforded
to us by rational intuition. But surely rational intuition works in some way (otherwise, a nasty skepticism and selfdefeat looms). And if it’s difficult for physicalism to explain this, so much the worse for physicalism, we say. But
even if you go in for quasi-experiential intellectual seemings, it’s still plausible that, and if it were false that 2+2=4,
we would have noticed that, just as, in the ordinary case of seeing an apple immediately before you, if the apple
weren’t there, you wouldn’t have believed that it was, because you would have noticed its absence. See also
Nicholas Sturgeon’s (1984, 65) response to a challenge to moral knowledge from Gilbert Harman. There, Sturgeon
says, “if a particular assumption is completely irrelevant to the explanation of a certain fact, then the fact would
have obtained, and we could have explained it just as well, even if the assumption had been false.” Though see
Joel Pust’s (2000, 240-243) criticisms.
Of course, talk of rational intuition often raises skeptical concerns about its fallibility, especially in light of
apparent disagreement, both of the intrapersonal variety (in the form of paradoxes), and the interpersonal variety
(in the form of disagreeing intuitions among philosophers). This concern is perhaps more acute when rational
intuition is said to be direct, unmediated, etc. How could such a power err? Readers wishing to explore this issue
may confer Bealer 1998, Bealer 2004, BonJour 1998, and, more recently, Bengson 2015, Christensen 2009,
Chudnoff 2016, Killoren 2010, and Mogensen 2017. We agree with Pust 2017, though, that “it can be difficult to
determine if another’s failure to have an intuition that p is epistemically significant, as they may have yet to really
grasp or consider the precise proposition at issue.” We think it a promising strategy worth exploring to diagnose
alleged cases of disagreement over intuitions (as well as alleged cases of errors in intuition) as cases of proposition
confusion in a Kripkean vein (cf. Kripke 1970, 143). Just as Kripke plausibly explained away the alleged intuitiveness
of “It’s possibly false that Hesperus = Phosphorus” as a mistaken description of the (true) intuition that “it’s
possibly false that the evening star = the morning star,” one might similarly explain away alleged cases of
disagreeing intuitions or intuitive error as actually involving a mistaken description of the relevant (and true)
intuition. Exploring this strategy fully is beyond the scope of this paper.
12
In the next section, we’ll turn to more difficult cases. These cases are so difficult that we call this
section…
Some Common Objections
We consider first the case of abductive knowledge. How could, for example, your belief that all
emeralds are green be explained by the relevant fact, when you’ve only seen a few emeralds? Here, we
believe that the field has already been plowed by previous philosophers. Carrie Jenkins (2006, 159),
drawing upon Goldman’s prior work (1984, 103), offers this explanation of abductive knowledge:
…[I]n genuinely knowledge-producing cases of inference to the best explanation, where p
explains q, q explains [that the agent believes that q], and [that the agent believes that q]
explains [that the agent believes that p], and all these explanatory links are standard, then one
can collapse the explanatory chain and simply say that p explains [that the agent believes that
p], even when talking to an outsider. For example, if I believe that all emeralds are green on the
grounds of an inference from known instances of green emeralds, it seems reasonable to say
that the fact that all emeralds are green is what explains my belief that they are. For it is what
explains the fact that all the emeralds I have seen so far have been green, and this is what
explains my general belief.
You may have qualms about endorsing transitivity with regard to explanation (or grounding), but notice
that this account of abductive knowledge doesn’t require anything so strong as that. It’s a limited,
qualified claim that, sometimes, under the right conditions, explanatory chains collapse, and that this is
one of those times with those right conditions. If Explanationism is true, then those “right conditions”
will have to do with difference-making, and which factors are strictly required for successful
explanation.
We move then to the truly difficult case of deductive knowledge, where previous work is not so
definitive. The prima facie worry is this: when I come to know something by way of deduction—when I
run an argument and draw a conclusion—don’t the premises explain my belief in the conclusion, rather
than the truth of the conclusion itself? Goldman (1984, 104-5) makes the following attempt to
incorporate deductive knowledge into Explanationism:
13
I know that a mouse is three feet from a four-foot flagpole with an owl on top, and I deduce that
the mouse is five feet from the owl. I then know this fact, although I have no explanation for the
mouse's being five feet from the owl, although this fact does not explain my belief. Appeal to
explanatory chains, however, allows us to assimilate this case to others within the scope of the
account. The relevant theorem of geometry is the common element in the relevant chains here.
First, the truth of that theorem explains why, given that the mouse is three feet from a four-foot
pole, it is five feet from the top, although it does not explain that fact simpliciter. Someone who
accepts the other two measurements but does not know why, given them, the mouse is five feet
from the owl, has only to be explained the theorem of geometry. Second, the truth of that
theorem helps to explain why mathematicians believe it to be true, which ultimately explains
my belief in it. My belief in the theorem in turn helps to explain my belief about the distance of
the mouse from the owl. Thus what helps to explain my belief in the fact in question also helps
to explain that fact, given the appropriate background conditions, which here include the other
distances. The case fits our account.
Goldman seems to use the following premises:
1. The Pythagorean theorem explains why, given that a mouse is three feet from a fourfoot-tall flagpole with an owl on top, then the mouse is five feet from the owl.
2. The Pythagorean theorem explains why mathematicians believe it.
3. The fact that mathematicians believe the Pythagorean theorem explains why I believe it.
4. My belief in the Pythagorean theorem helps explain my belief that the mouse is five feet
from the owl.
And Goldman wishes to derive this conclusion:
5. So, the Pythagorean theorem helps explain my belief that the mouse is five feet from
the owl, and also explains why, given that the mouse is three feet from a four foot
flagpole with an owl on top, the mouse is five feet from the owl.
If the explanatory chains collapse in the appropriate way, this conclusion does seem to follow: the
Pythagorean Theorem helps explain my belief, and helps explain why my belief is true. But what’s
missing in all this—we think, and Goldman seems to acknowledge—is any account of how the fact that
the owl is five feet from the mouse explains my belief that it is. Goldman shows, at most, how the
Pythagorean Theorem helps explain my belief and also its truth (given certain conditions); but he
14
doesn’t show how the truth of my belief explains why I believe it. And we took that to be the primary
desideratum of any Explanationist account of this case. Goldman seems content merely “to assimilate
this case to others within the scope of the account.” But we’d like something more: an Explanationist
account of deductive knowledge. So, we must look elsewhere.18
Carrie Jenkins (2006, 160-1) attempts to improve on Goldman, and considers a case of Brian
coming to know something by running the following argument:
6. If Neil comes to the party, he will get drunk.
7. Neil comes to the party.
8. So, he will get drunk
Brian knows (6) and (7), and believes (8) as a result of this argument. Does he know it, on
Explanationism? Jenkins says “yes,” and here’s her reasoning:
It’s only because Neil will get drunk that the two premises from which Brian reasons are true
together; that is to say, it is only because Neil will get drunk that Brian’s argument is sound…
[A]fter noting that competent reasoners like Brian will (generally) only use sound arguments, we
can explain the fact that Brian reasons as he does by citing the soundness of Brian’s
argument…Finally, we note that, uncontroversially, Brian’s reasoning as he did is what explains
his belief that Neil will get drunk. So now we know that the fact that Neil will get drunk explains
why Brian’s argument is sound, which explains Brian’s reasoning as he did, which in turn
explains why Brian believes that Neil will get drunk.
The first sentence is a little murky to us. Jenkins seems to be saying that (6) and (7) are true together,
and the argument is valid, because (8) is true. But, obviously, an argument can be unsound and still have
a true conclusion. (8) is no explanation of why “the two premises…are true together.” If anything, you
might have thought, it’s the other way around: (6) and (7) together explain the truth of (8).
We’re also puzzled by Jenkins’ claim that the fact that this argument is sound explains why
Brian, a competent reasoner, used it. After all, there are lots of sound arguments that Brian didn’t use.
18
An anonymous referee helpfully raised another concern, which you may well share: It is far from clear that the
Pythagorean Theorem construed as a proposition about pure Euclidean space can explain anything in the nonplatonic realm of matter. We take this to be another reason to leave Goldman’s proposal behind, and look
elsewhere for an Explanationist account of deductive knowledge.
15
Imagine before you a sea of sound arguments that Brian might have used. He happened to use this one
and not any other, and the fact that it’s sound won’t help us understand why, since they’re all sound.
We’d like to improve on the efforts from Goldman and Jenkins, like so. Deductive arguments are
a means by which we come to metaphorically “see” the truth of some proposition. Brian doesn’t believe
(8) because he believes (6), (7), and that (6) and (7) entail (8). Rather, those beliefs get him in a position
to see the truth of (8), and that is why he believes (8). It’s like an ordinary case of non-metaphorical
vision. Vision is fallible, since it might be a hallucination or illusion. But, so long as your visual system is
functioning well, it puts you in a position to see objects before you, and if one such is a chair, you
believe there’s a chair there because there’s a chair there. The fact that there’s a chair there is a crucial
part of the explanation of why you believe there is. Similarly, when you see the premises are true and
the inference valid, deduction helps you see the conclusion. And that’s because, in a sound argument,
the conclusion is already “there,” so to speak, in the premises. When one appreciates a sound
argument, one is not merely seeing the truth of the premises and the validity of the inference; one also
sees the truth of the conclusion thereby. In that case, you believe the conclusion because it’s true. This
is fallible, when the component beliefs or premises might be false or otherwise unknown. But, when
they’re true and you know it, then deduction positions you to see the truth of the conclusion.
So, ultimately, Brian does believe (8) because it’s true. His knowledge of the premises and
validity helped him appreciate that (8) is true. It’s like putting the pieces of a puzzle together, and then
seeing the picture. In this analogy, the premises and inference are the pieces, and the conclusion is the
picture. Now, when there’s a completed puzzle of a pangolin in front of you, and you believe there’s a
picture of a pangolin there, a crucial part of the explanation of why you believe there’s a picture of a
pangolin there is because there’s a picture of a pangolin there. In a similar way, when Brian knows those
premises together and grasps the entailment, he’s in position to see the truth of the conclusion. The
conclusion is now “part of the picture,” as it were, formed by the true premises and valid inference, and
is thereby directly appreciable. And so it’s correct that: Brian believes the conclusion because it’s true.
We may be misled into thinking otherwise by the way we standardly formalize arguments, with
conclusions on a separate line from the premises. If you worry this may be happening to you, remind
yourself that, with a valid argument, the conclusion is already “there” in the premises. When we
represent valid Aristotelian syllogisms using Venn diagrams, for example, after we draw the premises,
we needn’t add anything further to represent the conclusion: it’s already there. So, by appreciating the
truth of the premises, we are in a position to appreciate the truth of the conclusion. And, if we do, it can
16
be truly said that we believe the conclusion because it’s true. We think this is a satisfactory
Explanationist account of deductive knowledge, which can easily be adapted to other cases of deductive
knowledge.
We’ll turn now to inductive knowledge, especially of the future. On its face, such knowledge is a
real puzzle for Explanationism: Absent some tricky backwards causation, how could your current belief
that the sun will rise tomorrow be explained by the fact that the sun will rise tomorrow? Isn’t it rather
your current evidence that explains your current belief? Roger White (2010: 582-83) puts the concern
like so: “I’m justified in thinking the sun will rise tomorrow. But assuming it does, this fact plays no role
in explaining why I thought that it would.” To solve this puzzle, we think we can combine the abductive
and deductive solutions above. First, we’ll use the explanatory-chain-collapse strategy from Goldman
and Jenkins, like so:
9. You believe the sun rises every day because you've seen it rise many times in the past.
(SEEN explains BELIEVE)
10. The fact that that sun rises every day explains why you've seen it rise many times in the
past. (FACT explains SEEN)
11. So, the fact that the sun rises every day explains why you believe that the sun rises every
day. (So, FACT explains BELIEVE)
If every link in this chain is required for a complete explanation, then, according to Explanationism, you
know that the sun rises every day. Now, if you competently deduce that the sun will rise tomorrow from
you knowledge that the sun rises every day, it will be the case that: the fact that the sun will rise
tomorrow explains your belief that the sun will rise tomorrow. The known premise together with your
grasp of the implication puts you in a position to appreciate the truth of the conclusion. In that case, the
truth of the conclusion explains why you believe that it’s true. And that’s how inductive knowledge
works on Explanationism, we suggest.19
19
This same strategy may be employed to account for more mundane cases of knowledge of the future, for
example your knowing that you will get a raise next year, when your boss tells you so. So long as your boss is a
reliable testifier, i.e. so long as there’s a law-like connection between her testimony and the truth, it’s plausible
that a universal generalization like “if my boss asserts p, then p is true” explains (the positive track record of truthtelling you’ve observed from your boss, which explains) why you believe this is so. So, on Explanationism, you
know that this is so. And then, adding your knowledge of your boss’ assertion about your raise, you deduce that
you will get a raise next year, in such a way that the truth of this conclusion explains your belief in it, as described
above. So, it’s knowledge, according to Explanationism. One wrinkle: probably your boss is not so reliable that
there’s genuine entailment between her testimony and the truth. But, like much of our ordinary thought and
17
File the following objection under “Miscellaneous”:
The Case of the Epistemically Serendipitous Lesion. Suppose K suffers from a serious
abnormality—a brain lesion, let's say. This lesion wreaks havoc with K's noetic structure, causing
him to believe a variety of propositions, most of which are wildly false. It also causes him to
believe, however, that he is suffering from a brain lesion. K has no evidence at all that he is
abnormal in this way, thinking of his unusual beliefs as resulting from an engagingly original turn
of mind... surely K does not know that he is suffering from a brain lesion. He has no evidence of
any kind—sensory, memory, introspective, whatever—that he has such a lesion; his holding this
belief is, from a cognitive point of view, no more than a lucky (or unlucky) accident. (Plantinga
1993, 195)
The allegation here, adapting the case to target Explanationism, is that the fact that K has a brain lesion
plays a crucial role in K’s belief that he has a brain lesion, and so Explanationism must say K knows he
has a brain lesion, which is absurd. But we believe that this case can be handled in a similar way to
Faraci’s case of Eula’s random number generator, discussed above. Either the brain lesion is randomly
producing beliefs, or it isn’t. If it is randomly producing beliefs, then although K can’t know that he has a
brain lesion on this basis, Explanationism doesn’t have to say that he does. And that’s because, if the
brain lesion is operating genuinely randomly, then its operation does not explain why K ends up with the
belief that he has a brain lesion. We may have an explanation of why K believes something or other, but
no explanation of the particular belief that he has a brain lesion. That’s the nature of a truly random
process: there is no demystifying explanation of its operation. The story is simply this: K had a brain
lesion, and then—pop!—K ends up believing that he has a brain lesion. And there’s nothing demystifying
or explanatory about that pop.
On the other hand, suppose the brain lesion is not operating randomly when it tells K that he
has a brain lesion. Rather, the brain lesion tells K he has a brain lesion because he does have a brain
lesion. In a simple version of this case, when the brain lesion produces only the belief that there’s a
brain lesion, because it is a brain lesion, it looks like there is the right sort of explanatory connection for
Explanationism to say that this is a case of knowledge. But—and now buckle up, Internalists—it also
looks like that’s the right result: this is knowledge. The brain lesion is a good way for K to learn that he
has a brain lesion: the brain lesion “tells” K that he has a brain lesion because it’s true. And what more
reasoning, this indicative conditional may be nonmonotonic, and this a case of “defeasible reasoning.” That is,
there is likely a qualification commonly left implicit, to the effect that: unless something strange or anomalous
happens, if my boss asserts p, then p is true. If so, then in this case, what one really comes to know about the
future, according to Explanationism, is the more modest claim that: unless something strange or anomalous
happens, I will get a raise next year. But, on reflection, that sits well with us. We thank Bar Luzon for encouraging
us to think about cases like this.
18
could you ask from a brain lesion? This is merely a case of very unusual—though highly reliable—
testimony.
But the case is a little more complicated as Plantinga describes it. As Plantinga has it, the brain
lesion “wreaks havoc with K's noetic structure, causing him to believe a variety of propositions, most of
which are wildly false.” In this more complicated version of the case, although the brain lesion tells K
that he has a brain lesion not randomly but rather because it’s true, the brain lesion also tells K many
false things besides. And now it starts looking less like a case of knowledge. In fact, it starts looking more
like Fake Barn Country (cf. Alvin Goldman 1973). In Fake Barn Country, one’s eyes happen to alight on a
real barn amid a sea of fakes. In Chaotic Brain Lesion Country, one happens to receive a true belief from
a brain lesion selecting from a sea of falsehood.
What can the Explanationist say about Fake-Barn-style cases? Perhaps this. In the ordinary case
of seeing a barn, you believe there’s a barn before you because it looks like there’s a barn before you,
and—we continue, for a complete explanation—it looks like there’s a barn before you because there is a
barn before you. Explanationism gives its imprimatur as it should: this is knowledge. But, as the fake
barns proliferate, the second link in the explanatory chain becomes less plausible. Even if your eyes
happen to fall upon a real barn in a forest of fakes, we might begin to think it false that it looks like
there’s a barn before you because there is a barn before you. As the barn facades proliferate, a rival
explanation looms into view: that it looks like there’s a barn before you because you’re in a region full of
structures that look like barns. In other words, it becomes plausible to say that you believe there’s a
barn before you because it looks like there’s a barn before you, and it looks like there’s a barn before
you because you’re in Fake Barn Country. (In Fake Barn Country, it’s common to be appeared to barnly,
and more common the more barn facades there are. In that case, we can explain your belief by citing
the multitude of things that look like barns. There’s no need to cite this barn in particular to explain the
belief. Compare: when you drive over a road with very few nails, it’s true to say that your tire is flat
because you ran over that particular nail. But if you drive over a road with very many nails, we can
explain your flat tire by citing the multitude of nails. There’s no need to mention that nail in particular to
explain the flat.) And the fact that there’s a real barn before you plays no crucial role in that explanation.
So, on Explanationism, as the barns proliferate, it becomes less and less clear that you have knowledge.
At some point, perhaps, the knowledge disappears.20
20
In recent years, it’s become more common among epistemologists to express reservations about the claim that
one doesn’t know in Fake Barn Country. Perhaps this is because it’s not clear whether and to what extent it’s true
19
Something similar could be said about the more complicated version of Plantinga’s
serendipitous brain lesion. As the brain lesion produces more false-beliefs (which seem true from the
inside), the following explanation of K’s belief becomes more plausible: K believes that he has a brain
lesion because it seems true to him, and it seems true to him because he’s the victim of a brain lesion
that produces many true-seeming (but usually false) beliefs. And, in this case, the truth of K’s belief
plays no crucial role in the explanation of his believing it.21
Finally, let us close with a few words about Linda Zagzebski’s (1994) Paradox of the Analysis of
Knowledge, and how Explanationism answers it. Zagzebski argues that any view which allows “a small
degree of independence” between the justification and truth components of knowledge will suffer from
Gettier-style counterexamples. “As long as the truth is never assured by the conditions which make the
state justified,” she says, “there will be situations in which a false belief is justified” (ibid., 73). And,
“since justification does not guarantee truth, it is possible for there to be a break in the connection
between justification and truth, but for that connection to be regained by chance” (ibid. 67). Those are
Gettier cases.22 So the idea is that epistemologists are grappling with the following paradox:
(A) Knowledge = TB + X
(B) X does not entail TB. It’s possible to have X without TB. False but justified beliefs, for
example.
(C) At least some cases of X without TB can be “restored” to TB+X, without there being
knowledge.
that the alternative, knowledge-precluding explanation mentioned in this paragraph really does rise to
prominence, and become the explanation of the subject’s belief. We consider it a virtue of Explanationism that its
verdicts on Fake Barn Country cases seem to be as indeterminate as our intuitions about whether there’s
knowledge.
21
A similar thought might apply in the case of statistical inference, on Explanationism. One might know that a
lottery ticket probably will lose, on the basis of statistical evidence. But, since the fact that the ticket will lose
doesn’t figure into this evidence, one could not be said to know that the ticket will lose on this basis. This may help
solve the puzzle of why we’re reluctant to convict defendants on the basis of purely statistical evidence (cf.
Littlejohn 2017), but why we’re willing to convict on the basis of e.g. fallible eyewitness testimony (cf. Papineau
forthcoming). On Explanationism, eyewitness testimony can generate knowledge by providing a connection to the
truth, even if we’re uncertain that this connection holds. Statistical inference, on the other hand, provides at best a
connection to—and, so knowledge of—only probable truth. And evidently our judgments about knowledge reveal
that we much prefer a likely link to truth, rather than a link to likely truth.
22
See also Scott Sturgeon (1993) and Trenton Merricks (1995).
20
If you think that (A) there’s an analysis of knowledge to be had by adding a missing necessary condition
to truth and belief, and also that (B) this missing condition does not entail truth, and yet also that (C) any
such analysis will allow for Gettier cases, then you’ve contradicted yourself. That’s the problem.
Explanationism’s solution is to avoid Gettier cases by denying (B).23 On Explanationism, S knows
that p iff S believes that p because it’s true. Clearly, the right bijunct entails truth. So, Explanationism is
committed to denying (B). Zagzebski (ibid. 72) says “few philosophers have supported this view,” since it
means one must “give up the independence between the justification condition and the truth
condition.” But, justification (in the Internalist sense, at least) doesn’t figure into Explanationism’s
analysis. So, Explanationists are free to agree that there can be justified but false beliefs, in the sense of
“justification” that Zagzebski seems to have in mind. The Explanationist could add that, perhaps,
(Internalist) justification is just our best attempt to know whether the Explanationist condition is met. To
prove that it is. To prove that we know. But, knowing does not require knowing that one knows, or
being able to prove that one knows. And that’s because, in the final analysis, knowing is simply believing
something because it’s true.
Acknowledgments
For helpful conversations and comments, thanks to Julien Dutant, Gregory Gaboardi, Dan Greco, Alex
Grzankowski, Sydney Keough, Dan Korman, Clayton Littlejohn, Bar Luzon, Joel Mann, Barry Smith, and
audiences at the Institute of Philosophy at the University of London, as well as at St. Norbert College.
References
Achinstein, Peter (1983). The Nature of Explanation (New York: Oxford University Press)
Alspector-Kelly, Marc (2006). “Knowledge Externalism.” Pacific Philosophical Quarterly 87: 289–300
23
That is, on Explanationism, warrant entails truth. And this is plausible; cf. Andrew Moon (2012). This is an
explanation of why, in principle, there can be no Gettier cases on Explanationism. But, perhaps to confirm this
result, you might reflect on how Explanationism would handle a paradigm Gettier case. For example, Russell’s
Stopped Clock. Suppose your favorite and generally reliable clock has stopped this morning at 8:22 am. But, by the
sheerest coincidence, you happen to look at it this evening at 8:22 pm. Since the clock reads “8:22,” you form the
true belief that it’s 8:22 pm. And, given your history with this generally reliable clock, you have a high degree of
(Internalist) justification. But, no knowledge. Why is that, on Explanationism? Well, you do believe that it’s 8:22
because the clock reads “8:22.” But, crucially, it’s false that the clock reads “8:22” because it is 8:22. It reads “8:22”
because it was 8:22, when the clocked stopped. And, so, the truth of your belief does not figure crucially into the
explanation of why you hold the belief. So, Explanationism gives the right result here. And similar treatments could
be given of other Gettier cases.
21
Bealer, George (1998). “Intuition and the Autonomy of Philosophy,” in Michael DePaul & William
Ramsey (eds.), Rethinking Intuition: The Psychology of Intuition and Its Role in Philosophical Inquiry
(Rowman & Littlefield), 201-240
Bealer, George (2004). “The Origins of Modal Error,” dialectica 58(1):11-42
Becker, Kelly and Tim Black (2012). The Sensitivity Principle in Epistemology (Cambridge University Press)
Beddor, Bob and Carlotta Pavese (forthcoming). “Modal Virtue Epistemology,” Philosophy and
Phenomenological Research
Bengson, John (2015). “Grasping the Third Realm,” Oxford Studies in Epistemology, Volume 5, edited by
Tamar Szabo Gender and John Hawthorne (New York: Oxford University Press), 1-38
Black, Tim, and Peter Murphy (2007). “In Defense of Sensitivity,” Synthese 154 (1): 53–71
Bogardus, Tomas and Chad Marxen (2014). “Yes, Safety is in Danger,” Philosophia 42 (2): 321-334
BonJour, Laurence (1998). In Defense of Pure Reason (Cambridge: Cambridge University Press)
Broncano-Berrocal, Fernando (2014). “Is Safety in Danger?” Philosophia 42 (1):1-19
Christensen, Davn (2009). “Disagreement as Evidence: The Epistemology of Controversy”, Philosophy
Compass 4(5): 756–767
Chudnoff, Elijah (2016). Intuition (New York: Oxford University Press)
Clarke-Doane, Justin and Dan Baras (forthcoming). “Modal Security”
Clarke-Doane, Justin (2012). “Morality and Mathematics: The Evolutionary Challenge,” Ethics 122 (2):
313–40
Clarke-Doane, Justin (2014). “Moral Epistemology: The Mathematics Analogy.” Noûs 48 (2): 238–55
Clarke-Doane, Justin (2015). “Justification and Explanation in Mathematics and Morality,” in Russ ShaferLandau (ed.), Oxford Studies in Metaethics: Volume 1 (Oxford University Press)
Clarke-Doane, Justin (2016). “What Is the Benacerraf Problem?” In New Perspectives on the Philosophy
of Paul Benacerraf: Truth, Objects, Infinity, edited by F. Pataut, 17–43. Springer.
Clarke-Doane, Justin (forthcoming). “Debunking Arguments: Mathematics, Logic, and Modal Security.”
Robert Richards and Michael Ruse (eds.), Cambridge Handbook of Evolutionary Ethics (Cambridge:
Cambridge University Press)
DeRose, Keith (1995). “Solving the Skeptical Problem.” The Philosophical Review 104 (1): 1–52.
Dretske, Fred (1970). "Epistemic Operators", Journal of Philosophy 67:24, 1007-23
Dretske, Fred (1971). “Conclusive Reasons.” Australasian Journal of Philosophy 49 (1): 1–22.
Faraci, David (2019). “Groundwork for an explanationist account of epistemic coincidence,”
Philosopher’s Imprint 19(4): 1-26
22
Frankfurt, Harry (1969). “Alternate Possibilities and Moral Responsibility,” The Journal of Philosophy
66(23): 829-839
Goldman, Alan (1984). “An Explanatory Analysis of Knowledge,” American Philosophical Quarterly 21:
101-8
Goldman, Alan (1988). Empirical Knowledge (Berkeley: University of California Press)
Goldman, Alvin (1967) "Discrimination and Perceptual Knowledge," The Journal of Philosophy 73: 771-91
Hawthorne, John 2004. Knowledge and Lotteries. Oxford University Press
Ichikawa, Jonathan Jenkins (2011). “Quantifiers, Knowledge, and Counterfactuals.” Philosophy and
Phenomenological Research 82: 287–313
Jenkins, Carrie (2006). “Knowledge and Explanation,” Canadian Journal of Philosophy 36(2): 137-64
Killoren, David (2010). “Moral Intuitions, Reliability, and Disagreement,” Journal of Ethics and Social
Philosophy 4:1 (2010): 1-35
Korman, Daniel (2019). “Debunking Arguments,” Philosophy Compass 14(12): 1-17
Littlejohn, Clayton (2017). “Truth, knowledge, and the standard of proof in criminal Law,” Synthese 1–
34.
Luper-Foy, Steven (1984). “The Epistemic Predicament: Knowledge, Nozickian Tracking, and Scepticism,”
Australasian Journal of Philosophy 62 (1): 26–49
Merricks, Trenton (1995). “Warrant Entails Truth,” Philosophy and Phenomenological Research 55
(4):841-855
Mogensen, Andreas (2017). “Disagreements in Moral Intuition as Defeaters,” The Philosophical
Quarterly 67(267):282-302
Moon, Andrew (2012). “Warrant Does Entail Truth,” Synthese 184 (3):287-297
Neta, Ram (2002). “S Knows that P,” Nous 36: 663-81
Nozick, Robert (1981). Philosophical Explanations (Harvard University Press)
Papineau, David (forthcoming). “The Disvalue of Knowledge,” Synthese
Plantinga, Alvin (1993). Warrant: The Current Debate (Oxford University Press)
Pritchard, Duncan (2005). Epistemic Luck. Oxford University Press
Pritchard, Duncan (2007). “Anti-Luck Epistemology.” Synthese 158 (3): 277–98
Pritchard, Duncan (2009). “Safety-Based Epistemology: Whither Now?” Journal of Philosophical Research
34: 33–45.
Pust, Joel (2000). “Against Explanationist Skepticism regarding Philosophical Intuitions,” Philosophical
Studies 106(3): 227-258
23
Pust, Joel (2017). "Intuition", The Stanford Encyclopedia of Philosophy (Summer 2019 Edition), Edward
N. Zalta (ed.), URL = <https://plato.stanford.edu/archives/sum2019/entries/intuition/>.
Rieber, Steven (1998). “Skepticism and Contrastive Explanation,” Nous 32: 189-204
Roush, Sherrilyn (2005). Tracking Truth: Knowledge, Evidence and Science (Oxford University Press)
Sainsbury, Richard (1997). “Easy possibilities,” Philosophy and Phenomenological Research 57: 907-919
Sosa, Ernest (1999). “How to defeat opposition to Moore,” in J. Tomberlin (Ed.), Philosophical
Perspectives 13: Epistemology. Oxford: Blackwell. 141-54
Sosa, Ernest (2007). A Virtue Epistemology: Apt Belief and Reflective Knowledge. Vol. I. (Oxford
University Press)
Sosa, Ernest (2009). Reflective Knowledge: Apt Belief and Reflective Knowledge. Vol. II. (Oxford
University Press)
Strevens, Michael (2011). Depth (Harvard University Press)
Sturgeon, Nicholas (1984). 'Moral Explanations', in D. Copp and D. Zimmerman (eds.), Morality, Reason
and Truth (New Jersey: Rowman and Allenheld)
Sturgeon, Scott (1993). “The Gettier Problem,” Analysis 53 (3):156-164
Van Fraassen, Bas (1980). The Scientific Image (Oxford: Oxford University Press)
White, Roger (2010). “You Just Believe that Because...” Philosophical Perspectives 24: 573-615
Williamson, Timothy (2002). Knowledge and Its Limits (Oxford University Press)
Zagzebski, Linda (1994). “The Inescapability of Gettier Problems,” The Philosophical Quarterly 44(174):
65-73
24